(new in RDF4J 3.2)
RDF4J has (experimental) support for RDF* and SPARQL*.
RDF* and its companion SPARQL* are proposed extensions to the RDF and SPARQL standards (see Olaf Hartig’s position paper ) to provide a more convenient way to annotate RDF statements and to query such annotations. In essence, RDF* attempts to bridge the gap between the RDF world and the Property Graph world.
RDF4J support for these extensions currently (3.2.0) covers:
Note: these features are currently in the experimental/beta stage. While we’ll do our best to not make breaking changes in future releases unless necessary, we make no guarantees.
To support RDF* annotations, the core RDF model in RDF4J has been extended with a new type of Resource: the Triple
(not to be confused with the pre-existing Statement
, which is the representation of a “regular” RDF statement).
You can create Triple objects using a ValueFactory, and then use them as the subject (or object) of another statement, for example:
IRI bob = valueFactory.createIRI("http://example.org/bob");
Triple bobsAge = valueFactory.createTriple(bob, FOAF.AGE, valueFactory.createLiteral(23));
IRI certainty = valueFactory.createIRI("http://example.org/certainty");
Statement aboutBobsAge = valueFactory.createStatement(bobsAge, certainty, valueFactory.createLiteral(0.9));
The Statements
utility class offers several functions to easily transform Statements into Triples, vice versa:
IRI bob = valueFactory.createIRI("http://example.org/bob");
Triple bobsAge = valueFactory.createTriple(bob, FOAF.AGE, valueFactory.createLiteral(23));
Statement ageStatement = Statements.toStatement(bobsAge);
Triple backToTriple = Statements.toTriple(ageStatement);
RDF4J currently provides several Rio parser/writers for RDF*-enabled syntax formats: the Turtle* format, and the TriG* format. As the names suggest, both are extended versions of existing RDF format (Turtle and TriG, respectively). In addition, RDF4J’s binary RDF format parser has also been extended to be able to read and write RDF* data.
A Turtle* file that contains an annotation with a certainty score, on a statement saying “Bob’s age is 23”, would look like this:
@prefix foaf: <http://xmlns.com/foaf/0.1/> .
@prefix ex: <http://example.org/> .
<<ex:bob foaf:age 23>> ex:certainty 0.9 .
If we wish to read this data into an RDF4J Model object, we can do so using the Rio Turtle* parser:
Model model = Rio.parse(new FileInputStream("/path/to/file.ttls"), "", RDFFORMAT.TURTLESTAR);
Similarly, Rio can be used to write RDF* models to file:
Rio.write(model, new FileOuputStream("/path/to/file.ttls"), "", RDFFORMAT.TURTLESTAR);
Note: not every store can handle RDF* data. Attempting to upload an RDF* model directly to a Repository that does not support it will result in errors.
The RDF4J MemoryStore accepts RDF* data. You can add the RDF* model we created above directly to an in-memory Repository:
try(RepositoryConnection conn = repo.getConnection()) {
conn.add(model);
}
You can query this data via the Repository API, like any “normal” RDF data. For example:
try(RepositoryConnection conn = repo.getConnection()) {
RepositoryResult<Statement> result = conn.getStatements(null, null, null);
result.forEach(System.out::println);
}
will output:
<<<http://example.org/bob> <http://xmlns.com/foaf/0.1/age> 23>> <http://example.org/certainty> 0.9
and of course the subject triple can be inspected in code as well:
try(RepositoryConnection conn = repo.getConnection()) {
RepositoryResult<Statement> result = conn.getStatements(null, null, null);
Statement st = result.next();
Triple rdfStarTriple = (Triple)st.getSubject();
System.out.println(rdfStarTriple.getSubject()); // will output http://example.org/bob
}
When querying a store that contains RDF* data, even regular SPARQL queries may include RDF* triples as a result. To make it possible to serialize such query results (e.g. for network transfer), the SPARQL/JSON, SPARQL/XML, Binary and TSV query result formats have all been extended to handle RDF* triples as possible value bindings.
The default SPARQL 1.1 Query Results JSON format has been extended as in the following example:
{
"head" : {
"vars" : [
"a",
"b",
"c"
]
},
"results" : {
"bindings": [
{ "a" : {
"type" : "triple",
"value" : {
"s" : {
"type" : "uri",
"value" : "http://example.org/bob"
},
"p" : {
"type" : "uri",
"value" : "http://xmlns.com/foaf/0.1/age"
},
"o" : {
"datatype" : "http://www.w3.org/2001/XMLSchema#integer",
"type" : "literal",
"value" : "23"
}
}
},
"b": {
"type": "uri",
"value": "http://example.org/certainty"
},
"c" : {
"datatype" : "http://www.w3.org/2001/XMLSchema#decimal",
"type" : "literal",
"value" : "0.9"
}
}
]
}
}
RDF4J introduces a new, custom content type application/x-sparqlstar-results+json to specifically request this extended content. By default, if a client requests the standard content type (application/sparql-results+json or application/json), the response serializer will instead “compress” the RDF* triple into a single IRI, using Base64 encoding:
{
"head" : {
"vars" : [
"a",
"b",
"c"
]
},
"results" : {
"bindings": [
{ "a" : {
"type" : "uri",
"value" : "urn:rdf4j:triple:PDw8aHR0cDovL2V4YW1wbGUub3JnL2JvYj4gPGh0dHA6Ly94bWxucy5jb20vZm9hZi8wLjEvYWdlPiAiMjMiXl48aHR0cDovL3d3dy53My5vcmcvMjAwMS9YTUxTY2hlbWEjaW50ZWdlcj4-Pg=="
},
"b": {
"type": "uri",
"value": "http://example.org/certainty"
},
"c" : {
"datatype" : "http://www.w3.org/2001/XMLSchema#decimal",
"type" : "literal",
"value" : "0.9"
}
}
]
}
}
This ensures that by default, an RDF4J-based endpoint will always use standards-compliant serialization to avoid breaking clients that have not yet been updated.
It is possible to override this behavior by explicitly configuring the writer to not encode RDF* triples. Setting the BasicWriterSetting ENCODE_RDF_STAR
to false will ensure that even when requesting the standard content type, the serializer will use the extended syntax.
The default SPARQL Query Results XML format has been extended as in the following example:
<?xml version='1.0' encoding='UTF-8'?>
<sparql xmlns='http://www.w3.org/2005/sparql-results#'>
<head>
<variable name='a'/>
<variable name='b'/>
<variable name='c'/>
</head>
<results>
<result>
<binding name='a'>
<triple>
<subject>
<uri>http://example.org/bob</uri>
</subject>
<predicate>
<uri>http://xmlns.com/foaf/0.1/age</uri>
</predicate>
<object>
<literal datatype='http://www.w3.org/2001/XMLSchema#integer'>23</literal>
</object>
</triple>
</binding>
<binding name='b'>
<uri>http://example.org/certainty</uri>
</binding>
<binding name='c'>
<literal datatype='http://www.w3.org/2001/XMLSchema#decimal'>0.9</literal>
</binding>
</result>
</results>
</sparql>
As with the extended SPARQL Results JSON format, this format is sent by default only when specifically requesting a custom content type: application/x-sparqlstar-results+xml. And as with the JSON format, the XML writer can be reconfigured to also send the extended format on requesting the regular SPARQL/XML content type, by setting the BasicWriterSetting ENCODE_RDF_STAR
to false.
The SPARQL 1.1 Query Results TSV format has been extended as follows:
?a ?b ?c
<<<http://example.org/bob> <http://xmlns.com/foaf/0.1/age> 23>> <http://example.org/certainty> 0.9
The SPARQL engine in RDF4J has been extended to allow for SPARQL* queries. Executing a SPARQL* query relies on the underlying store supporting RDF* data storage.
SPARQL* allows accessing the RDF* triple patterns directly in the query. For example, after you have uploaded the above simple RDF* model to a MemoryStore, you can execute a query like this:
PREFIX foaf: <http://xmlns.com/foaf/0.1/>
PREFIX ex: <http://example.org/>
SELECT ?p ?a ?c WHERE {
<<?p foaf:age ?a>> ex:certainty ?c .
}
The result will be:
?p ?a ?c
ex:bob 23 0.9
RDF4J offers functions to convert between RDF* data and regular RDF. In the converted regular RDF, the RDF* triple is replaced with a reified statement using the RDF Reification vocabulary. For example:
<<ex:bob foaf:age 23>> ex:certainty 0.9 .
becomes:
_:node1 a rdf:Statement;
rdf:subject ex:bob ;
rdf:predicate foaf:age ;
rdf:object 23 ;
ex:certainty 0.9 .
You can find the the conversion functions in the Models
utility class. There are several variants, but the simplest form just takes a Model containing RDF* data and creates a new Model containing the same data, but with all RDF* annotation converted to reified statements:
Model rdfStarModel; // model containing RDF\* annotations
Model convertedModel = Models.convertRDFStarToReification(rdfStarModel);
Likewise, you can convert back:
Model reificiationModel; // model containing RDF reification statements
Model rdfStarModel = Models.convertReificiationtoRDFStar(reificiationModel);
Note: since the RDF* functionality is currently in experimental stage, it is possible that the names or precise signatures of these functions will be changed in a future release.
In the current discussions around RDF*, two “modes” for working with RDF* have been identified. Simply put, these two modes are:
<<ex:bob foaf:age 23>> ex:certainty 0.9 . implicitly also asserts the statement ex:bob foaf:age 23..<<ex:bob foaf:age 23>> ex:certainty 0.9 does not automatically imply that the statement ex:bob foaf:age 23 is present - the ground fact and the annotation are seen as independent entities, and either one can exist without the other.Although RDF4J makes no assumptions about the mode used at the level of its core interfaces, the default implementations in RDF4J’s own triplestore implementations currently work only in SA mode.
Table of Contents